Adaptive routing behavior with real-time information under multiple travel objectives

Abstract

Real-time information about traffic conditions is becoming widely available through various media and connected-vehicle technology. In such conditions, travelers have better knowledge about the system and adapt as the system evolves dynamically during their travel. Drivers may change routes during their travel in order to optimize their own objective of travel. Various travel objectives are captured in mathematical models via disutility functions. The focus of this research was to study the behavior of travelers with multiple trip objectives when they are provided real-time information, and assess their ability to determine “optimal” routing policies, compared to exact solutions based on the online shortest path problem. A web-based experiment was carried out to simulate a traffic network with limited information provision. The decision strategies of participants were analyzed and compared to a variety of decision policies established in the literature – optimal, greedy, and a priori – and a general model to describe the observed travelers’ decision strategies was calibrated from over 40,000 decision points extracted from the collected data. Apart from trip objective, other factors such as relative position in the network and experience gained were found to influence user decisions. This research is a step towards calibrating equilibrium models for adaptive behavior with multiple user classes.

Keywords

• Online shortest path
• Adaptive routing
• Real-time information
• Routing games
• Driver behavior